X370 Taichi and 1700X
Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=5309
Printed Date: 10 Dec 2024 at 9:40pm Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: X370 Taichi and 1700X
Posted By: Ken429
Subject: X370 Taichi and 1700X
Date Posted: 05 Jun 2017 at 5:35pm
" rel="nofollow - I just assembled my first AMD based system! It consists of: X370 Taichi MB BIOS Version 230 X1700 Ryzen Processor running at 3.8GHz EVGA Memory 2X8GB DDR4 3000 1.35V Cas Latency 15 MSI GTX 560 TI Video Adapter MasterLiquid Pro 280 AOI Samsung 250GB 960 EVO NVMe M.2 Boot Device Crucial 500 GB MX200 Seasonic 760XP Power Supply Corsair 750D Case Windows 10 Pro ver. 1703
I'm almost a very happy camper. The AMD system transcodes a test BluRay to MP4 in 41% less time than a 4790K running @ 4.6GHz. That's about what I expected based upon all the benchmark stuff on the internet.
Several things still aren't quite right:
ASRock's "Restart to UEFI" does not work - it boots right back to W10.
The Z97 Extreme/4790K system with the Samsung 960 EVO NVMe device boots to the Post Screen in ~6 seconds. The X370 Taichi/1700X system takes ~9 seconds to boot to the Post Screen. Both systems have the Samsung device in the Ultra M.2 slot. It almost seems like the Taichi BIOS is having trouble with something. I'm about ready to make an "old" Sandisk SSD the boot drive!
The CPU temperature is a bit of mystery - under full load on all threads for an extended period HWInfo64 says the maximum temperatures reached were 77.6 (Tctl) and 57.6 (Tcle) (the 20C anomaly?). Seems pretty low considering the CPU is clocked at 38 and using a Vcore of 1.28V. Either I have a very good chip or I'm not understanding what is going on. Most of the people posting on the Internet are using much higher voltages to get to 3.8GHz. This thing is very stable and seems like it wants to go faster @ 1.28V? I haven't spent much time trying to find what Vcore setting is the minimum for 3.8GHz but I've gone from 1.36V to 1.30V and now 1.28V and the CPU temperatures, if I can believe them, don't change that much?!
|
Replies:
Posted By: Ken429
Date Posted: 08 Jun 2017 at 5:58pm
" rel="nofollow - In my original Post I said the Z97 Extreme6/4790K BIOS posted from a cold boot in ~6 seconds and I mistakenly (Or the X370 Taichi/1700X is getting slower) since it takes ~19 seconds to boot to the first BIOS screen from a cold start. Something ain't right! Is it just me or is this a known issue?
|
Posted By: Dogmifier
Date Posted: 08 Jun 2017 at 7:07pm
" rel="nofollow - Ken, question
did you set it to "Fast" or "Ultra fast" boot? If so, I don't think it will go into UEFI until you reset CMOS and try again.
As for the boot time, I think the AMD Post times are a bit slower than Intels...this may change as the platform matures...
The CPU temp I think stems from software not reading things correctly...tho again, I could be wrong...my temps and voltages jump all over using any software that monitors them
I don't have a taichi, but a gaming K4
------------- X99X Fata1ity 5930k 16GB Crucial 1200W Corsair evga GTX 1080 Samsung 950 Pro
|
Posted By: Ken429
Date Posted: 08 Jun 2017 at 7:17pm
" rel="nofollow - I just edited my post while you were answering my original post. Taking ~19 seconds to get to the first BIOS screen is a symptom of a problem? I'll try the Fast Boot and see what happens.
|
Posted By: Dogmifier
Date Posted: 08 Jun 2017 at 7:22pm
Ken429 wrote:
" rel="nofollow - I just edited my post while you were answering my original post. Taking ~19 seconds to get to the first BIOS screen is a symptom of a problem? I'll try the Fast Boot and see what happens. |
I would say the post taking 19 seconds is (not correct, but correct for now) right...post is slow on these machines right now as far as I know..
maybe parsecs or wardog has better insight than I do..
------------- X99X Fata1ity 5930k 16GB Crucial 1200W Corsair evga GTX 1080 Samsung 950 Pro
|
Posted By: Ken429
Date Posted: 08 Jun 2017 at 7:34pm
" rel="nofollow - Dogmifier, I started to set the BIOS to Ultafast boot and then read the description in the BIOS and got cold feet. It seems to say that once set to Ultrafast the only way to get back into the BIOS setup is to Clear CMOS. Since I'm still playing with the BIOS settings I don't think I want to go there?!
|
Posted By: Dogmifier
Date Posted: 08 Jun 2017 at 7:48pm
Ken429 wrote:
" rel="nofollow - Dogmifier, I started to set the BIOS to Ultafast boot and then read the description in the BIOS and got cold feet. It seems to say that once set to Ultrafast the only way to get back into the BIOS setup is to Clear CMOS. Since I'm still playing with the BIOS settings I don't think I want to go there?! |
Yeah, that's what it does...I was asking if you had that set, not to set it....
sorry for the misunderstanding...
Not sure why you're not going into UEFI if that's not set.
------------- X99X Fata1ity 5930k 16GB Crucial 1200W Corsair evga GTX 1080 Samsung 950 Pro
|
Posted By: Ken429
Date Posted: 08 Jun 2017 at 7:59pm
" rel="nofollow - Dogmifier,
I am able to enter the BIOS? AND I want to be able to do that since I'm still playing with the multiplier and Vcore looking to see what this thing will do. I don't know if 3800 MHz at 1.28125V is good or not. Still looking for some input from others with this processor and possibly M/B.
|
Posted By: datonyb
Date Posted: 08 Jun 2017 at 10:30pm
x370 taichi (bios version 2.0) r7 1700 (non X ) OVERCLOCKED IN P STATES TO 3.9 GHZ all cores voltage set at fixed 1.306v (in p states) LLC set at level 2 SOC voltage set at 1.0v LLC for soc ,set fixed at level 2 m2 nvme sansung evo 960 as boot drive sata 3 ssd samsung evo 850 as data/games drives 2 x 8gb tridentz ram running at 3200 mhtz from day one amd boot training turned OFF
to be honest i dont find the boot time an issue its circa 5-10 seconds to windows log in screen its quite fine for me it gives me time to hit the del key if i want to go into bios or sit down and place my ciggies on desk before log in screen
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: parsec
Date Posted: 08 Jun 2017 at 11:44pm
Ken429 wrote:
" rel="nofollow - I just assembled my first AMD based system! It consists of: X370 Taichi MB BIOS Version 230 X1700 Ryzen Processor running at 3.8GHz EVGA Memory 2X8GB DDR4 3000 1.35V Cas Latency 15 MSI GTX 560 TI Video Adapter MasterLiquid Pro 280 AOI Samsung 250GB 960 EVO NVMe M.2 Boot Device Crucial 500 GB MX200 Seasonic 760XP Power Supply Corsair 750D Case Windows 10 Pro ver. 1703
I'm almost a very happy camper. The AMD system transcodes a test BluRay to MP4 in 41% less time than a 4790K running @ 4.6GHz. That's about what I expected based upon all the benchmark stuff on the internet.
Several things still aren't quite right:
ASRock's "Restart to UEFI" does not work - it boots right back to W10.
The Z97 Extreme/4790K system with the Samsung 960 EVO NVMe device boots to the Post Screen in ~6 seconds. The X370 Taichi/1700X system takes ~9 seconds to boot to the Post Screen. Both systems have the Samsung device in the Ultra M.2 slot. It almost seems like the Taichi BIOS is having trouble with something. I'm about ready to make an "old" Sandisk SSD the boot drive!
The CPU temperature is a bit of mystery - under full load on all threads for an extended period HWInfo64 says the maximum temperatures reached were 77.6 (Tctl) and 57.6 (Tcle) (the 20C anomaly?). Seems pretty low considering the CPU is clocked at 38 and using a Vcore of 1.28V. Either I have a very good chip or I'm not understanding what is going on. Most of the people posting on the Internet are using much higher voltages to get to 3.8GHz. This thing is very stable and seems like it wants to go faster @ 1.28V? I haven't spent much time trying to find what Vcore setting is the minimum for 3.8GHz but I've gone from 1.36V to 1.30V and now 1.28V and the CPU temperatures, if I can believe them, don't change that much?!
|
The 19 second POST time sounds like you are checking the Task Manager Startup tab "Last BIOS time". If that is right, that is exactly what I get with my ASRock X370 Killer SLI/ac board with a 1700X. Completely normal so far with Ryzen.
Ryzen POST time is slow compared to an Intel system. As datonyb said, you can disable AMD Advanced Boot Training to speed up POST time, but in some cases your memory may not be able to reach high speeds with that option disabled. Ryzen does not support a memory speed of 3000, so what is your memory at now, 2933?
Using the Ultra Fast setting of Fast Boot won't change the POST time very much. I use it and the difference is only a few seconds. You'll get a shorter startup time just disabling AMD Advanced Boot Training. It won't get the startup time down to Intel levels, but it is improved. Don't forget this is all POST time, nothing to do with Windows booting. Z97 boards have very short POST times for some reason, shorter than Z170. That might be due to the change in using the chipset resources for the M.2 slots in Z170, instead of the CPU's PCIe lanes in Z97. Ryzen is using the CPU's PCIe lanes for any M.2 slots that are PCIe 3.0 x4. The M2_2 slot on the X370 Taichi is PCIe 2.0 x4 through the chipset.
Personally, using a 960 EVO as the OS drive, once I hear the single POST Ok beep, the OS boots almost as fast as an Intel PC, I don't see any spinning dots when Windows is starting. That of course is from a cold boot, when the Win 10 fast startup feature is active. It takes a little longer on a restart, but not terrible. Ryzen's small file 4K read performance is not as good as Intel's so far, it suffers if any CPU power saving features are active. The same it true for Intel PCs, but so far Ryzen is a bit slower under the same circumstances.
POST time on Intel board's like a Z170 were known to be slower when using NVMe SSDs as the OS drive. Z270 PC are faster for some reason. Windows boot time has clearly hit a wall, since it is not much faster with NVMe SSDs than SATA SSDs.
You are not understanding what you must do to use the Ultra Fast setting, and how to get into the UEFI/BIOS when you use it.
Your Windows installation must be full UEFI booting, with CSM disabled or at least the Launch Storage OpROM Policy option set to UEFI Only, with CSM enabled. That needs to be done when you install Windows, and will not work with Windows 7. You use Win 10, so that is not an issue.
When you install Windows, you must select in the boot order the entry for the Windows installation media the entry that is, "UEFI: <device name>", which signals the Windows installer to partition the OS drive as GPT, and creates four partitions.
When using the Ultra Fast setting, you must use the ASRock Restart to UEFI program that you run in Windows. Otherwise you cannot get into the UEFI/BIOS the usual ways. I've had problems with this program for the first time on my Ryzen PC, it fails to work with a memory OC for some reason. Always works perfectly with my Intel PCs, been using it for years. You have the same problem I see. Restart to UEFI actually uses a feature added to Windows starting with Win 8, that restarts the PC right into the UEFI/BIOS interface. Why it fails with Ryzen, I don't know.
The two CPU temperatures you see, Tctl and Tdie are programmed into a Ryzen processor by AMD. You are correct the difference (offset) is exactly 20° C. The point of those two temperatures, according to AMD, is apparently to control the stock AM4 CPU cooler's fan speed, given the differences between eight core Ryzen 7 processors, and Ryzen 5 processors. Tdie is the correct temperature to monitor with a Ryzen 7 CPU.
3.8GHz is not a high OC, as I'm sure you know. Normally the VCore needs increase above 3.8GHz. But it seems you have a good CPU.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: Ken429
Date Posted: 09 Jun 2017 at 1:43am
" rel="nofollow - [URL=][/URL][URL=][/URL]parsec,
Thanks for the information and I guess agreement that that is the way AMD and Ryzen operate. I'm a long time Intel fan but could not resist the 16 threads for transcoding stuff.
The ~19 seconds I was quoting varies by +3 seconds depending on the "time of day" or how the system feels at the time! When I look at the Task Manger "Last BIOS" time" is says pretty consistently 14.3 seconds. The ~19+ that I was quoting was according to my watch and counting 1 second, 2 seconds, etc. Not very official but close to the actual time it takes from when I hit the Power Button until the first BIOS screen appears - meaning the one that shows the option to enter the BIOS etc.
The one thing that I may have screwed up on the W10 fresh installation on a brand new SSD was I ended up with an MBR file structure on the Samsung 960. In the past I've always cloned a new SSD and/or formatted it as GPT prior to installing Windows. I did not think this was a problem, is it? If so I can reinstall W10 and force a GPT structure.
I do not fully understand "Fast Boot" and have no intention of using it.
But...are you saying that the reason the ASRock "Restart to UEFI" program is not working for me on this machine is because the boot drive is formatted as MBR? If that's the case I'll happily reinstall W10.
Sorry, I did not read your post close enough, I will try down clocking the EVGA 3000 memory to the default 2400 setting and see if that helps the ASRock "Restart to UEFI". I will also return the BIOS settings to the Defaults and see if that is where my original ~9 seconds came from.
|
Posted By: Optik
Date Posted: 09 Jun 2017 at 2:18am
Ryzen need longer to post than other CPUs, that's normal. I'm happy that my system running very smooth and stable. Maybe 10sec and I can login into win10.
------------- Taichi X370 AMD R7 1700X Arctic Liquid FreeZer 240 G.Skill Flare X 3200 cl 14@2933 PowerColer RX480 Phanteks EnTHoo Luxe BeQuiet Dark Power Pro 11 550W
|
Posted By: Ken429
Date Posted: 09 Jun 2017 at 4:35am
" rel="nofollow - [URL=][/URL]parsec,
Ok, I'm not crazy. I put everything in the BIOS back to the defaults including the memory and I'm back to ~9 seconds to go from pushing the power button to the first BIOS post screen! The ASRock "Restart to the UEFI" also works as advertised.
So...wait for ASRock/AMD to fix whatever is going on when the memory is running at Profile 1 speeds (for my EVGA 3000 > 2933)? Curious and lazy, if I leave the memory at default and up the multiplier does the problem(s) 20 seconds to post and Restart to UEFI restart to Windows come back? In other words is there a cut off point for memory speed and OC'ing that I can go to before the issues return?
By the way Task Manager still shows Last BIOS time: 14.3 seconds??
|
Posted By: datonyb
Date Posted: 09 Jun 2017 at 4:52am
im assuming the extra time is the bios trying to set the faster ram speed
eg stock ddr4 speed and settings it'll just start to boot due to knowing it safe settings but when setting the ram overclock its checking to make sure its going to boot safely and thus the extra time
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: parsec
Date Posted: 09 Jun 2017 at 9:59am
Ken429 wrote:
" rel="nofollow - [URL=][/URL][URL=][/URL]parsec,
Thanks for the information and I guess agreement that that is the way AMD and Ryzen operate. I'm a long time Intel fan but could not resist the 16 threads for transcoding stuff.
The ~19 seconds I was quoting varies by +3 seconds depending on the "time of day" or how the system feels at the time! When I look at the Task Manger "Last BIOS" time" is says pretty consistently 14.3 seconds. The ~19+ that I was quoting was according to my watch and counting 1 second, 2 seconds, etc. Not very official but close to the actual time it takes from when I hit the Power Button until the first BIOS screen appears - meaning the one that shows the option to enter the BIOS etc.
The one thing that I may have screwed up on the W10 fresh installation on a brand new SSD was I ended up with an MBR file structure on the Samsung 960. In the past I've always cloned a new SSD and/or formatted it as GPT prior to installing Windows. I did not think this was a problem, is it? If so I can reinstall W10 and force a GPT structure.
I do not fully understand "Fast Boot" and have no intention of using it.
But...are you saying that the reason the ASRock "Restart to UEFI" program is not working for me on this machine is because the boot drive is formatted as MBR? If that's the case I'll happily reinstall W10.
Sorry, I did not read your post close enough, I will try down clocking the EVGA 3000 memory to the default 2400 setting and see if that helps the ASRock "Restart to UEFI". I will also return the BIOS settings to the Defaults and see if that is where my original ~9 seconds came from.
|
No, it's not MBR partitioning that causes Restart to UEFI (RtU) to fail. I have my 960 EVO GPT partitioned, and Restart to UEFI fails when my memory is over clocked.
You said in a post after the one I quoted that after a UEFI/BIOS reset, that RtU worked. Isn't that with your memory at default/stock speed? It did work, I just checked your other post.
If RtU works with your memory over clocked (anything over 2667), great! I did not test RtU with the latest UEFI version for my AM4 board, but with other versions it is ignored, and Windows just boots after running RtU. First PC it has ever failed working with. I use RtU all the time, since it is such a convenience, there should be a button on the board or IO panel to do the same thing.
Forgot to explain the Fast Boot option. It really is "fast POST", meaning it reduces the POST time. That can have side affects. A PC's startup time is POST time plus Windows loading and running time. "Boot" time generally means both together, but really refers to loading the OS.
My 19 second BIOS time is caused by using AMD's RAID, which adds a few seconds to POST.
With CSM enabled, the UEFI firmware in our boards runs in emulated legacy BIOS mode, from 1998. 16 bit addressing and only 1MB of DRAM available, which today is a joke. But we are held back by legacy hardware still being used, since it is not UEFI compatible. I've been UEFI booting since my ASRock Z77 Extreme4 board, with Windows 8, using SATA SSDs.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: Ken429
Date Posted: 09 Jun 2017 at 7:53pm
I tried leaving the OC at 38 and dropping the Memory speed to default and everything works OK. Meaning it boots to the first BIOS screen in ~9 seconds and Restart to UEFI works too.
Just in case you want to know the effect of the slower memory speed on my test transcode to MP4, the time went from ~64 minutes (2933) to ~66 minutes (2133). Not as bad as all the memory posts would have me believing but they are more game oriented not just brute computing.
|
Posted By: datonyb
Date Posted: 09 Jun 2017 at 8:42pm
bios version v2.40 is out today that should help with the memory
im just testing overclocking settings now
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: Ken429
Date Posted: 11 Jun 2017 at 2:42am
" rel="nofollow - Installed BIOS Version 2.40 and cannot find anything that improved. Still takes ~19 seconds to boot and Restart to UEFI does not work. Can somebody give me a clue on what got better?
Are people using the GSkill Memory with the Samsung Chips having the same issue?
One other thing that is a little perplexing - the rated speed of the EVGA 3000 memory is supposed to be CL 15. Yet CPU-Z says it got changed to CL 16. Does that mean that the system does not like the memory @ CL 15?
|
Posted By: datonyb
Date Posted: 11 Jun 2017 at 9:30pm
i too have cl15 ram and it changes the first setting to 16 no matter what i do even back in bios it still has 15 written but in windows its cl16
if i set the all to 14 it will stay at 14
but then wont COLD boot at cl14 at 3200 cl15 xmp (all be it with one changed to 16) it boots fine and has done on all bios from 1.6 onwards
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: parsec
Date Posted: 11 Jun 2017 at 10:38pm
Ken429 wrote:
" rel="nofollow - Installed BIOS Version 2.40 and cannot find anything that improved. Still takes ~19 seconds to boot and Restart to UEFI does not work. Can somebody give me a clue on what got better?
Are people using the GSkill Memory with the Samsung Chips having the same issue?
One other thing that is a little perplexing - the rated speed of the EVGA 3000 memory is supposed to be CL 15. Yet CPU-Z says it got changed to CL 16. Does that mean that the system does not like the memory @ CL 15?
|
The X370 Taichi has the AGESA 1.0.0.6 from AMD that was supposed to help with memory compatibility and speed. The results so far have been mixed, depending upon the memory being used, users report better, worse, or unchanged memory speeds and support. Some people got lucky, others not.
AGESA 1.0.0.6 brings more memory speed and timing options, and other rather unique Ryzen options that remain a mystery to me at this time. But all the options in the world don't help if we don't know what to do with them. For example, the ODT options are frustrating, since I've never seen an ODT spec for any model of memory.
My X370 Killer SLI/ac board, with 16GB of G.SKILL 3200 FlareX memory (made for Ryzen, Samsung chips) with a new, similar UEFI update, leaves me so far mostly unchanged in memory operation. I'm still at 2933 14 14 14 34 1T (which is not bad), but 3200 still won't work for me, POST fails. I tried the new 3066 speed at 16 16 16 36, but that has poor stability in Windows, browser tabs crashing but not the browser program entirely (never saw that before.) That is with the SOC voltage at 1.2V too. So yes, I have the same type of result.
Restart to UEFI is still failing for me too, when the memory is at 2933. That is strange since that program should be based on a Windows feature introduced in Windows 8, and I've never seen it fail on my many ASRock boards.
What got better may be a factor of our intervention in tuning memory options, so what I call auto-magic improvements (fixes with no effort on our part) again are a matter of luck.
If you set the timing(s) to 15 (not Auto) and it was changed to 16, then yes the system prefers 16. Or it became "trained" to that value. Relaxed timings can cause memory to work at slightly higher speeds. Ryzen does not support a speed of exactly 3000, does 2933 work for you? We do have a 3066 speed option now, I assume that fails for you? 3066 is a slight memory OC for your memory, so if many of us cannot get our memory to run at the rated (but still an OC relative to the Ryzen memory controller spec) OC speed, then a true memory OC will usually be more difficult.
For an experiment, you can try setting all the memory timings to Auto (system selected), and try 2933 or 3066. DRAM Voltage at 1.35V, SoC voltage at 1.2V.
------------- http://valid.x86.fr/48rujh" rel="nofollow">
|
Posted By: Ken429
Date Posted: 13 Jun 2017 at 7:24pm
" rel="nofollow - parsec,
I finally got my nerve up to try some other memory settings.
I tried 2400 versus the 2133 and the system hung on a BIOS code of "A6".
I then tried 3066 and the system would not boot and hung in a BIOS "OF" loop. I had to shut the system off, count to 10 and Clear CMOS to get it to boot. I left the VDDCR_SOC voltage set @ Auto. Based on A-Tuning System Information Display it said it's running @ 1.112V.
There used to be a list of the BIOS codes in the manual that came with the M/B. Apparently ASRock is trying to save a penny or two and left them out. Is there a place on the Web to download the codes for the X370 Taichi M/B?
I'm more that a little confused on why the system won't run @ 2400. I suppose I should take your suggestion and increase the SOC voltage to 1.2V but I'm not sure anything will help this thing since it seems to only want to run @ 2133 or 2933 settings. It must be the need to 33!! Maybe BIOS version 2.5 will help.
I tried upping the VDDCR_SOC voltage to your recommended 1.200V. Did that and A-Tuning says it at 1.208V. The system runs ok with added voltage and the Memory set to 2933. Went back into the BIOS and set the Memory to 3066 and the BIOS hung with an "OF" code. Had to Clear CMOS twice to get the system back. The first Clear BIOS hung with a code of "54". Sure would be nice to no what the codes are telling me!
All of this is going on with the EVGA 3000 CL15 memory. Seems like serious Ryzen users like the GSkill memory that uses Samsung chips. Should I give it a try or will I run into the same roadblock.
|
Posted By: datonyb
Date Posted: 15 Jun 2017 at 1:14am
ken if you can find it the tridentz f4 c3200c15d-16gtz kits work on xmp 3200
but they are circa 150 bucks and i doubt you would actually notice the difference in windows from 2933 speed
------------- [url=https://valid.x86.fr/jpg250][/url]
3800X, powercolor reddevil vega64, gskill tridentz3866, taichix370, evga750watt gold
|
Posted By: Ken429
Date Posted: 15 Jun 2017 at 6:48pm
" rel="nofollow - datonyb,
I did not ask the question correctly. Will a Ryzen 1700X/Taichi M/B combination with Trident Z memory kit rated @ 3200 or 3600 run any faster than 2933?
|
Posted By: Xaltar
Date Posted: 15 Jun 2017 at 8:35pm
Anything past 2666 offers only a negligible performance improvement. From 2933 to 3200 on my setup yields only about a 3 - 5 point boost to performance in Cinebench R15 and maybe 1 - 2 FPS in popular games like Rise of the Tomb Raider using low settings @720p on my GTX 960. As soon as I raise the settings to 1080p and high the difference is gone (GPU becomes the bottleneck).
My setup:
Ryzen 5 1600x 16gb Corsair Vengeance LPX DDR4 3600 (CMK16GX4M2B3600C18) @3200 on defaults with XMP on GTX 960 2gb
From what I have seen in my testing so far, you are better off saving money on the RAM and getting something like a 2666 kit and hoping to overclock it to 2933. A lot of people with 3200 and above kits are struggling to hit 3200 and beyond even with very expensive kits. Its a lot of effort for very little gain. Unless you are running a GTX 1080 or better and plan to game at 1080p @120hz or above you will not see a significant difference between 2666 and 3200+.
AGESA 1.0.0.6 has helped for some when it comes to reaching higher speeds but it isn't worth the money, headache and time in my opinion.
-------------
|
Posted By: AlbinoRhino
Date Posted: 15 Jun 2017 at 8:51pm
" rel="nofollow -
Xaltar wrote:
Anything past 2666 offers only a?negligible performance improvement. From 2933 to 3200 on my setup yields only about a 3 - 5 point boost to performance in Cinebench R15 and maybe 1 - 2 FPS in popular games like Rise of the Tomb Raider using low settings @720p on my GTX 960. As soon as I raise the settings to 1080p and high the difference is gone (GPU becomes the bottleneck).?
My setup:
Ryzen 5 1600x 16gb Corsair Vengeance LPX DDR4 3600 (<span style=": rgb43, 43, 43;">CMK16GX4M2B3600C18</span>) @3200 on defaults with XMP on GTX 960 2gb
From what I have seen in my testing so far, you are better off saving money on the RAM and getting something like a 2666 kit and hoping to overclock it to 2933. A lot of people with 3200 and above kits are struggling to hit 3200 and beyond even with very expensive kits. Its a lot of effort for very little gain. Unless you are running a GTX 1080 or better and plan to game at 1080p @120hz or above you will not see a significant difference between 2666 and 3200+.
AGESA 1.0.0.6 has helped for some when it comes to reaching higher speeds but it isn't worth the money, headache and time in my opinion.? |
I found this to be true as well. I paid far less for my CL16 Ram for basically the same performance as the CL14 that everyone is recommending for Ryzen.
I've been running 2933 since day 1 of my system build and with my 1700X the performance is superb.
Programs load quickly for productivity and gaming on my Ultrawide is all done at ultra graphics settings with no performance drops.
Edit: I have the GTX 1080 GPU.
|
Posted By: Ken429
Date Posted: 15 Jun 2017 at 9:19pm
" rel="nofollow - So... there are Ryzen systems with 3200 or 3600 memory that are running above 2933. I am not a gamer. The only reason I built the Ryzen system was to run Handbrake and it does a fine job at that as I stated above. The one thing I did determine that going from 2133 to 2933 yielded a 2-3 percent performance improvement with Handbrake. 2133 to 2933=+800/2133=+37.5% 2933 to 3200=+267/2933=+9.1% 2933 to 3600=+667/2933=+22.7%
You're correct, going from 2933 to even 3600 if it were possible on the Ryzen 1700X CPU probably yields minimal Handbrake performance improvement, maybe 1 or 2 percent. But it's good to know that there are people running above 2933 with the right memory?
My system is normal at 2133. But at 2933 it takes and extra ~10 seconds to boot and the ASRock "Restart to UEFI" does not work, it boots right back to Windows. Do you have the same issues at 3200?
|
Posted By: Xaltar
Date Posted: 15 Jun 2017 at 9:29pm
" rel="nofollow - Thanks for sharing your findings AlbinoRhino and Ken429
I think the biggest issue is that a lot of people take review figures as being more significant than they really are. A reviewer will deliberately force the bottleneck onto the component being tested so it is possible to compare similar parts. This is almost never a reflection of real world use. Who is going to use a Pentium G4560 with a Titan XP for example?
People need to stop reading/watching reviews and taking that information out of context. Most reviewers will state why they tested the way they did. I write reviews myself and have been working in the industry since the days of the 486 and this hasn't changed. Reviews show you a situation in which the product on review is pushed to it's limits and compared to it's competition under the same conditions. The odds that a casual user would even have the requisite hardware to do this are slim, add to that the specific settings needed and you will almost never see the same behavior in real world use.
The other factor, especially with games, is how little difference even 20fps difference makes once you are talking 60+ fps, maybe 100+ fps with 100hz+ monitors today. The simple fact is, users will not notice the difference between 210fps and even 300fps. So when we see DDR4 3200 increasing gaming performance by 10-20fps (depending on title and settings) with a 1080ti or Titan XP when compared to DDR4 2133 it really isn't that significant.
That said, Ryzen does benefit from faster RAM and that can be reasonably significant in certain situations making it worth it to spring for 2666 or better RAM. Once we pass 2666, barring good deals/rebates we enter diminishing returns. It makes absolutely no sense to get CL14 3200 RAM and pair it with an RX 460 or GTX 1050ti, get cheap RAM, overclock it as much as you can and pick up an RX 570 or GTX 1060. Now if you are going all out, then by all means have some fun with it and pick up the super fast RAM with the silly tight timings and have fun fine tuning and tweaking your system, this is the joy of being an enthusiast after all.
Component choices are based on budget and the rules shift the higher up you go (budget wise).
-------------
|
|